Mobile neuroimaging has long promised to take cognitive neuroscience out of the lab and into the world. The challenge isn’t conceptual – it’s practical. Wearable EEG and fNIRS hardware exist, but dedicated software for running cognitive tasks on a mobile device has been missing. Keeping multiple data streams in sync adds another layer of difficulty.
A new study by Rokowski and colleagues at Villanova University introduces WearCAAT – the Wearable Cognitive Assessment and Augmentation Toolkit. It provides the first formal validation of this mobile neuroimaging platform in healthy adults.
The Setup: Wearable Sensors, One Tablet
57 college-aged participants wore a full-head hybrid cap. It housed a combined wireless fNIRS-EEG system covering frontal, temporal, and parietal regions. The fNIRS system was a NIRx system with 51 channels – 43 long-distance and 8 short channels for skin artifact removal. EEG was recorded via mbt’s Smarting PRO mobile system. All streams were synced through the lab streaming layer (LSL), which handled clock alignment and jitter removal.
Participants used WearCAAT on an Apple iPad Pro. The app ran all 11 cognitive tasks – about 1.5 hours of data – in a single sitting. Zero participants dropped out. Zero app crashes occurred. No data was lost or corrupted across any stream. For a mobile neuroimaging platform in its first formal test, that is a meaningful result.
The WearCAAT data pipeline: the app streams metadata, stimuli, and behavioral markers from the iPad wirelessly via LSL to a laptop. fNIRS data travels via USB to the software; EEG streams wirelessly via Bluetooth. Both signals are saved together as a synchronized .xdf file. This pipeline is what makes concurrent mobile neuroimaging practical in a portable, non-lab setting.The Task: Visual Oddball on a Touchscreen
For validation, the authors chose the visual oddball paradigm. It has well-documented EEG and fNIRS correlates that can serve as clear benchmarks. Participants saw letter strings on the iPad screen: frequent standard stimuli (OOOOO) and rare target stimuli (XXXXX). They responded by tapping one of two buttons at the bottom of the screen. Each stimulus appeared for 0.5 seconds. A 1.5-second blank interval followed each one. The task ran for 4 minutes, flanked by 10-second baseline periods.
That runtime was a deliberate design choice. WearCAAT is built for multi-task batteries. Each task adds to a broader cognitive profile rather than acting as a standalone long session. The key question is then pointed: can valid neural signals come from a task that runs in a fifth of the normal time? This is a real challenge in any mobile neuroimaging context.
Behavioral Responses Confirmed the Oddball Effect
Participants responded more slowly to target stimuli than to standard ones – the classic RT gap of the oddball paradigm. Mean accuracy exceeded 94% for targets and 99% for standards. The RT difference was robust across the full sample.
One note for future mobile neuroimaging work: participants used one finger to tap between two buttons. Standard desktop setups provide two dedicated physical controls. That physical shift likely inflated target RTs. The authors flag it as fixable through clear setup instructions.
P300 Amplitude Was Clear and Consistent
Grand-average ERP waveforms at Cz and Pz for all participants. Target responses are in red, standard in blue; shaded regions show the standard error of the mean. At both sites, the target trace shows a clear positive deflection peaking in the 250–400 ms range – the P300 – while the standard trace stays considerably lower. The clean separation between conditions at both midline sites confirms that the mobile setup captured genuine attentional processing.All 53 analyzed participants showed higher P300 amplitude for target stimuli than for standard ones. This held at both the parietal midline (Pz) and central midline (Cz). Peak latencies at both sites fell within the 250–400 ms window the published literature reports.
The waveforms were noisier than those from full-length desktop protocols. Fewer trials means less averaging. Residual variability survives more easily into the grand mean. This is a known trade-off in short mobile neuroimaging designs. The authors expect longer task runs to produce smoother waveforms.
At Cz, the target trace rises clearly above the standard in the 250–400 ms window. The shaded standard error bands separate cleanly between conditions. Pz tells the same story – a positive deflection for targets peaking well ahead of the standard, with little overlap past 300 ms.
Prefrontal HbO Increased Only for Target Stimuli
Mean HbO activations in the right frontal ROI and individual channels. The target response (red) follows the classic hemodynamic response shape — a slow rise peaking around 9 seconds, then a return to baseline. The standard response (blue) stays flat. The right-hand panels show the same trend across each individual fNIRS channel in the ROI, confirming the spatial consistency of the cognitive engagement signal.The fNIRS data backed up the EEG findings from a hemodynamic angle. Oxygenated hemoglobin (HbO) in the right prefrontal cortex rose in response to target stimuli. It followed a canonical hemodynamic response curve, peaking around 9 seconds post-stimulus before returning to baseline. Standard stimuli produced no such increase.
The pattern held across every channel in the right frontal region of interest, not just the aggregate. Right PFC lateralization of the HbO response matches previous fNIRS oddball studies using computer-based tasks. Seeing it replicate in a mobile neuroimaging setup – on a 4-minute iPad task – is the cross-validation the field needs. Pocket laboratory designs won’t go mainstream without it.
The aggregate HbO trace shows the target following a clean hemodynamic curve. The standard stays flat. The individual channel panels show the same trend across every channel in the ROI. That spread rules out a noise spike in one part of the array. It supports a genuine attention-related signal.
What This Means for the Future of Mobile Neuroimaging
WearCAAT produced behavioral, EEG-based, and hemodynamic outcomes that matched published norms. That convergence – across three independent measures, on a mobile neuroimaging platform, in 4 minutes – is the study’s core result.
The platform collected concurrent EEG and fNIRS with zero data loss across 57 sessions. Full validation of the remaining 10 tasks is ongoing. But the visual oddball data make a strong case that pocket laboratory designs are technically ready. The next step is taking mobile neuroimaging where labs can’t go.
Read the original paper by Rokowski et al., linked here.
Curious if this setup is right for your study?
Our team works with researchers across neuroscience, sports science and clinical applications - we'll help you figure out if mobile EEG fits your study design.
